1

Letter of Intent

for the Compressed Baryonic Matter

Experiment at the

Future Accelerator Facility in Darmstadt

The CBM collaboration

Darmstadt, January 2004

2

Abstract The CBM Collaboration proposes to build a dedicated heavy-ion experiment to investigate the properties of highly compressed baryonic matter as it is produced in nucleus-nucleus collisions at the future accelerator facility in Darmstadt. Our goal is to explore the QCD phase diagram in the region of moderate temperatures but very high baryon densities. The envisaged research program includes the study of key questions of QCD like confinement, chiral symmetry restoration and the nuclear equation of state at high densities. The most promising diagnostic probes are vector mesons decaying into dilepton pairs, strangeness and charm. We intend to perform comprehensive measurements of hadrons, electrons and photons created in collisions of heavy nuclei. CBM will be a fixed target experiment which covers a large fraction of the populated phase space. The major experimental challenge is posed by the extremely high reaction rates of up to 107 events/second. These conditions require unprecedented detector performances concerning speed and radiation hardness. The detector layout comprises a high resolution Silicon Tracking System in a magnetic dipole field for particle momentum and vertex determination, Ring Imaging Cherenkov Detectors and Transition Radiation Detectors for the identification of electrons, an array of Resistive Plate Chambers for hadron identification via TOF measurements, and an electromagnetic calorimeter for the identification of electrons, photons and muons. The detector signals are processed by a high-speed data acquisition and trigger system.

3

The CBM Collaboration (Group leaders underlined) Bergen, Norway, Department of Physics, University of Bergen D. Röhrich Bucharest, Romania, National Institute for Physics and Nuclear Engineering C. Aiftimiei, V. Catanescu, M. Ciobanu, D. Moisa, M. Petris, A. Petrovici, M. Petrovici, A. Raduta, G. Stoicea Budapest, Hungary, Eötvös University F. Deak, R. Izsak, A. Kiss Budapest, Hungary, KFKI E. Denes, Z. Fodor, J. Kecskemeti, Cs. Soos, T. Kiss, G. Vesztergombi Coimbra, Portugal, LIP P. Fonte, R. Marques, A. Policarpo Cyprus University, Nikosia, Cyprus J. Mousa, H. Tsertos Darmstadt, Germany, GSI M. Al-Turany, A. Andronic, H. Appelshäuser, E. Badura, E. Berdermann, D. Bertini, P. Braun-Munzinger, M. Deveaux, J. Eschke, H. Fleming, V. Friese, C. Garabatos, R. Holzmann, P. Koczon, W. Koenig, B. Kolb, Y. Leifels, C. Lippmann, W.F.J. Müller, A. Schüttauf, K. Schwarz, P. Senger, R. Simon, J. Stroth Dubna, Russia, JINR-LHE A. Baldin, E. Baldina, V. Chepurnov, G. Cheremukhina, S. Chernenko, O. Fateev, A. Ierusalimov, A. Malakhov, E. Matyushevsky, V. Pechenov, O. Rogachevski, L. Smykov, Yu. Zanevsky, V. Zrjuev Dubna, Russia, JINR-LPP Ju. Gousakov, N. Grigalashvili, G. Kekelidze, D. Peshekhonov, V. Peshekhonov, V. Livinski, S. Mishin, K. Viriasov, Ju. Zlobin Dubna, Russia, JINR-LIT P. Akishin, E. Akishina, I. Alexandrov, S. Baginyan, S. Dmitrievskii, Valeri Ivanov, V. Ivanov, E. Litvinenko, G. Ososkov, A. Raportirenko, A. Sosnin, P. Zrelov Frankfurt, Germany, Institut für Kernphysik, Universität Frankfurt C. Müntz, H. Ströbele

4

Heidelberg, Germany, 2. Physikalisches Institut, Universität Heidelberg M. L. Benabderahmane, E. Cordier, N. Herrmann, A. Mangiarotti Heidelberg, Germany, Kirchhoff-Institut für Physik, Universität Heidelberg D. Atanasov, I. Kisel, V. Lindenstruth, G. Torralba Katowice, Poland, University of Silesia A. Grzeszczuk, S. Kowalski, M. Krauze, E. Stephan, W. Zipper Kharkov, Ukraine, University of Kharkov M. Rekalo Kiev, Ukraine, Shevshenko University O. Bezshyyko, I. Kadenko, D. Kresan, V. Plujko, V. Shevshenko Krakow, Poland, Jagiellonian University J. Brzychczyk, J. Cibor, R. Karabowicz, Z. Majka, P. Staszel, P. Szostak, W. Walus Mannheim, Germany, Inst. of Computer Engineering, Universität Mannheim K.-H. Brenner, U. Brüning, U. Ehrbächer, P. Fischer, J. Gläß, P. Haspel, R. Männer, C. Steinle, A. Wurz Marburg, Germany, Fachbereich Physik, Universität Marburg B. Kohlmeyer, C. Schindler Moscow, Russia, Institute for Nuclear Research M. Golubeva, F. Guber, O. Karavichev, T. Karavicheva, E. Karpechev, A. Kurepin, A. Maevskaia, I. Peshenichnov, V. Rasin, A. Reshetin, V.Tiflov, N. Topilskaya Moscow, Russia, ITEP A. Akindinov, A. Arefiev, Y. Grishuk, A. Golutvin, S. Kiselev, I. Korolko, T. Kvaracheliya, V. Maiatski , S. Malyshev, A. Martemiyanov, E. Melnikov, P. Polozov, A. Smirnitskiy, V. Stolin, K. Voloshin, B. Zagreev, Smolyankin laboratory: Y. Grishkin, A. Lebedev, N. Rabin, V. Smolyankin, A. Zhilin Moscow, Russia, SINP, Moscow State University N. Baranova, G. Bashindjagyan, D. Karmanov, M. Korolev, M. Merkin, A. Voronin Moscow, Russia, Kurchatov Institute A. Kazantsev, V. Manko, I. Yushmanov

5

Münster, Germany, Institut für Kernphysik, Universität Münster D. Bucher, M. Hoppe, K. Reygers, A. Wilk, J. Wessels Obninsk, Russia, Obninsk State University of Atomic Energy V.Galkin, A.Golovin, D.Rizhikov, V.Saveliev, M.Zaboudko Prag, Czech Republic, Technical University V. Petracek, L. Skoda Protvino, Russia, IHEP V. Ammosov, M. Bogolyubsky, V. Dyatchenko, Yu. Kharlov, V. Khmelnikov, A. Kuznetsov, V. Obraztsov, B. Polishchuk, V. Rykalin, A. Ryazantsev, S. Sadovsky, A. Semak, V. Shelikhov, Yu. Sviridov, V. Zaets Pusan, Korea, Pusan National University D.-S. Kim, J.-Y. Kim, I.-K. Yoo Rez, Czech Republic, Czech Academy of Sciences D. Adamova, A. Kugler, J. Novotny, P. Tlusty Rossendorf, Germany, FZR, Institut für Kern- und Hadronenphysik F. Dohrmann, E. Grosse, K. Heidel, B. Kämpfer, R. Kotte, L. Naumann Santiago de la Compostela, Spain, University M. Angeles Lopez, D. Belver, J. Garzon, F. Gomez, D. Gonzalez Seoul, Korea, Korea University B. Hong, Y.J. Kim, K.S. Sim St. Petersburg, Russia, Khlopin Radium Institute (KRI) M. Chubarov, V. Karasev, S. Loshaev, Y. Murin, V. Plujschev, E. Seleznev St. Petersburg, Russia, CKBM A. Bolonin, V. Dobulevich, S. Igolkin, G. Karasev, G. Petrova, A. Svischev, M. Tkachev, V. Varava, A. Vasiliev, A. Vorobeva, V. Yaritzin St. Petersburg, Russia, PNPI V.Ivanov, A. Khanzadeev, A.Nadtochii, Yu.Riabov, V. Samsonov, D.Seliverstov, M.Zhalov St. Petersburg, Russia, St. Petersburg State Polytechnic University Y. Berdnikov, V. Grebenschikov, M. Ryzhinskiy

6

Strasbourg, France, IN2P3-CNRS/ULP (IRes) A. Besson, G. Gaycken, S. Heini, F. Rami, H. Souffi-Kebbati, M. Winter Warszawa, Poland, University, Nuclear Physics Division M. Kirejczyk, T. Matulewicz, B. Sikora, K. Siwek-Wilczynska, K. Wisniewski Zagreb, Croatia, Rudjer Bošković Institute Z. Basrak, R. Čaplar, M. Dželalija, I. Gašparić, M. Kiš Contact: Peter Senger, p.senger@gsi.de The base of this document is the Conceptual Design Report (CDR) on “An International Accelerator Facility for Beams of Ions and Antiprotons”. The CDR includes the motivation for and the physics programme of the Compressed Baryonic Matter experiment, the experimental observables, a preliminary sketch of the experimental setup, the required properties of the detector subsystems and first results of feasibility studies. This Letter of Intent contains an update of the detector concept, recent results of feasibility studies and detector developments, and the definition of working packages and responsibilities.

7

Contents 1. Introduction and overview 2. Detector subsystems. 2.1. Superconducting dipole magnet 2.2. Silicon Tracking System (STS) 2.3. Ring Imaging Cherenkov Detectors (RICH) 2.4. Transition Radiation Detectors (TRD) 2.5. Resistive Plate Chambers for Time-of-flight measurements (RPC) 2.6. Electromagnetic calorimeter (ECAL) 2.7. Diamond pixel detector for TOF start signal 3. Trigger, data acquisition and event reconstruction 3.1. Trigger an DAQ design considerations and developments 3.2. Event reconstruction 4. Physics performance 4.1. Detector simulation software package 4.2. Hadron identification by TOF 4.3. Low-mass vector meson identification 4.4. D meson detection 4.5. J/Ψ detection 5. Experimental programme and beam time estimate 5.1. Beam requirements 5.2. Experimental programme 5.3. J/Ψ meson measurements 5.4. D meson measurements 5.5. Low-mass vector meson measurements 5.6. Hyperons, antiprotons, fluctuations, hadron flow, exotica 5.7. Medium energy research programme 6. Responsibilities, schedule, radiation environment 6.1. Responsibilities 6.2. Funds 6.3. Working packages 6.4. Time schedule 6.5. Radiation environment Appendix 1: Detector studies A1.1. Experimental conditions A1.2. The Silicon Tracking System A1.3. Transition Radiation Detector development A1.4. Studies for a Ring Imaging Cherenkov detector A1.5. Recent achievements in R&D on Resistive Plate Chambers A1.6. The electromagnetic calorimeter system A1.7. CVD diamond strip detectors for TOF measurements and beam monitoring Appendix 2: The superconducting dipole magnet Appendix 3: HADES @ SIS100/300

8

1. Introduction and overview In this Letter of Intent we propose to design and to build a detector system in order to study nucleus-nucleus collisions at the future accelerator facility in Darmstadt [1]. The scientific goal of the research program is to explore the phase diagram of strongly interacting matter in the region of highest baryon densities. This approach is complementary to the activities at RHIC (Brookhaven) and ALICE (CERN-LHC) which concentrate on the region of high temperatures and very low net baryon densities. The territory of dense baryonic matter accessible in heavy-ion collisions is located between the line of chemical freeze-out and the hadronic/partonic phase boundary, as indicated by the hatched area in figure 1.1. New states of matter beyond the deconfinement and chiral transition at high net baryon densities and moderate temperatures may be within the reach of the experiment. The proposed experimental programme includes:

• the study of in-medium properties of hadrons, • the search for the chiral and deconfinement phase transition at high baryon densities, • the search for the critical point of strongly interacting matter, • the study of the nuclear equation-of-state of baryonic matter at high densities (as it

exists in the interior of neutron stars), • the search for new states of matter at highest baryon densities.

These topics are of fundamental interest both for astrophysics [2] and Quantum Chromodynamics (QCD) at high baryon densities [3].

Figure 1.1: The phase diagram of strongly interacting matter. The red symbols represent freeze-out points obtained with a statistical model analysis of particle ratios measured in heavy-ion collisions [4,5,6]. The pink curve refers to a calculation of the chemical freeze-out which occurs at a constant baryon density (baryons + antibaryons) of ρB =0.75 ρ0 (with ρ0=0.16 fm-3). The blue curve represents the phase boundary as obtained with a QCD lattice calculation [7] with a “critical point” (blue dot) at T = (160 ± 3.5) MeV and µB = (725 ± 35) MeV (ρB ≈ 3ρ0 ). In the region of the blue circle the baryon density is ρB ≈ 0.038 fm-3 ≈ 0.24 ρ0 (“dilute hadronic medium”). The corresponding value for the red circle is ρB ≈ 1 fm-3 ≈ 6.2 ρ0 (“dense baryonic medium”). The hatched area marks the region of matter at high baryon densities.

9

The experiment aims at a comprehensive study of hadrons, electrons and photons emitted in heavy-ion collisions. The key observables include:

• Low-mass vector mesons (ρ, ω, φ) decaying into e-e+ pairs. Via the measurement of penetrating probes one can extract the in-medium spectral function of mesons and, hence, obtain information on the possible restoration of chiral symmetry in dense baryonic matter.

• Hidden and open charm (charmonium, D mesons). The measurement of charmed mesons at threshold beam energies will shed light on the in-medium production processes and on the properties of highly compressed strongly-interacting matter.

• (Multi-) Strange baryons (Λ, Ξ, Ω). The yields and phase-space distributions of baryons containing (newly created) strange quarks are expected to be sensitive to the early and dense stage of the collision.

• Global features like the collective flow and critical event by event fluctuations. These observables contain information on the nuclear equation-of-state at high densities and on the existence of a critical point, and, hence, on the location and the order of the deconfinement phase transition.

• Direct photons from first collisions and thermal photons from the dense and hot fireball.

• Exotica like pentaquarks, bound kaonic systems, strange clusters, precursor effects of a color super-conducting phase, etc. The creation of compressed baryonic matter possibly will lead to unexpected phenomena.

The main experimental objective is the measurement of extremely rare signals in an environment typical for heavy-ion collisions. This requires the collection of a huge number of events which can only be obtained by very high reaction rates and long data taking periods. We aim at reaction rates of up to 10 MHz (minimum bias) which corresponds to a beam intensity of 109 beam particles per second on a 1 % interaction target (see Appendix A1.1). The rare signals are embedded in a large background of charged particles. For example, a central Au+Au collision at 25 AGeV produces about 1000 charged particles which in addition create a similar amount of secondaries via reactions in the target and detector materials. Hence, the experimental challenge is to measure about 1010 charged particles per second which are focused into a small forward cone, and to identify rare particles.

The experimental setup has to fulfill the following requirements:

• identification of electrons which requires a pion suppression factor in the order of 104, • identification of hadrons with large acceptance, • determination of the primary and secondary vertices (accuracy ≈ 50 µm), • high granularity of the detectors, • fast detector response and read-out, • very small detector dead time, • high-speed trigger and data acquisition, • radiation hard detectors and electronics, • tolerance towards delta-electrons.

Figure 1.2 depicts the present layout of the CBM experimental setup. As compared to the first layout as shown in the CDR, the RICH detector is split into two and the TRD is split into three subdetectors in order to improve the tracking capabilities of the setup. Additional tracking

10

stations might be needed in order to reconstruct the particle trajectories from target to the last detector. Full tracking simulations are being performed to clarify this question. The first RICH serves for the identification of electrons from the decays of low-mass vector mesons. The second RICH is optional. It might be used for the separation of high energy pions and kaons in order to improve the identification of high-energy D mesons. Detailed simulations will be performed to prove the viability of such a RICH detector at high particle multiplicities. An electromagnetic calorimeter (ECAL) is added to the layout as it was shown in the CDR in order to improve the pion suppression capability of the setup and to be able to measure direct photons and neutral mesons decaying into photons. The ECAL offers also the possibility to identify muons.

The CBM setup is optimized for heavy-ion collisions in the beam energy range from about 8 to 45 AGeV. Experiments on dilepton production at beam energies from 2 to about 8 AGeV could be carried out with the HADES spectrometer if installed in front of the CBM target. Simulations are performed to optimize the performance of HADES at higher beam energies (see Appendix 3).

Figure 1.2: Geometry of the CBM experiment. Inside the dipole magnet gap are the target and a 7 plane Silicon Tracking System (STS) consisting of pixel and strip detectors. The Ring Imaging Cherenkov detectors (RICH) have different radiators for the detection of electrons (first RICH) and mesons (second RICH). The Transition Radiation Detector (TRD) arrays measure electrons with momenta above 1 GeV (γ = 2000). The TOF stop detector consists of Resistive Plate Chambers (RPC). The electromagnetic calorimeter (ECAL) measures electrons, photons and muons. References [1] http://www.gsi.de/GSI-Future/cdr/ [2] F. Weber, J. Phys. G: Nucl. Part. Phys. 27 (2001) 465 [3] F. Wilczek, Physics Today 53 (2000) 22; hep-ph/0003183 [4] P. Braun-Munzinger et al., Phys. Lett. B 465 (1999) 15 [5] J. Cleymans and K. Redlich, Phys. Rev. Lett. 81 (1998) 5284 [6] R. Stock, Phys. Lett. B 456 (1999) 277 [7] Z. Fodor and S.D. Katz, Phys. Lett. B 534 (2002) 87

11

2. Detector subsystems

2.1. Superconducting Dipole Magnet

The momentum of charged particles is derived from their track curvature in the magnetic dipole field in which the target and the Silicon Tracking System (STS) are placed. The goal is to build a dipole magnet with large aperture (1x1 m2) and low stray field. The field integral has to be at least 1 Tm to achieve a sufficient momentum resolution and to deflect most of the abundantly produced δ electrons. A possible solution is presented in Appendix A2.

2.2. Silicon Tracking System (STS)

The experimental concept is to track charged particles directly after the target with a compact detector system. Several layers of silicon pixel and strip detectors are placed inside the large gap of the dipole magnet. The tracking system has the following tasks:

• momentum measurement for charged tracks with ∆p/p < 1% (0.5 < p/(GeV/c) < 4), • vertex determination with a resolution of better than 50 µm, • efficient recognition of electron pairs from π0 decays (γ conversion, Dalitz).

A particular aspect of the system is to achieve high track reconstruction efficiency in a high track density environment also for incomplete tracks, i.e. those originating from regions within the vertex tracker, or curved tracks which leave the acceptance before passing the last tracking layer. It is therefore essential to get redundant information from the detector system. The reconstruction of track pairs originating from secondary vertices, in particular γ conversions in the detector material, will be essential to reduce the combinatorial background in the electron pair spectrum.

The vertex resolution required for open charm detection is about 50 µm (see chapter 4.3). To achieve this, we aim at a position resolution of a single vertex tracker layer below 10 µm and a respective material budget below 10-3 X0. Both requirements could be fulfilled by Monolithic Active Pixel Sensors (MAPS). However, radiation hardness and readout speed of existing MAPS prototypes are still insufficient for CBM requirements. Therefore, R&D concentrates on these two issues (see Appendix A1.2). Another solution could be next generation hybrid pixel sensors. The sensors developed for LHC have the advantage of being sufficiently radiation hard and achieving high readout speed already today, but, in contrast to the MAPS, their material budget is substantially larger due to the fact that two chips are surface mounted by ball bonding technique. Here the technological progress of thinning and bonding techniques could remove this disadvantage in the future.

The bulk area of the tracking stations will be covered by silicon strip detectors with matched strip geometry. The detailed configuration of this tracking detectors is subject of detailed simulations based on realistic detector response and fully developed tracking algorithms.

2.3. Ring Imaging Cherenkov Detectors (RICH)

The first RICH detector provides identification of electrons and suppression of pions in the momentum range of electrons from low-mass vector-meson decays. The performance of the RICH is presently studied in simulations in order to optimize the refraction index of the

12

radiator gas, the geometry and the material of the mirrors and the geometry and granularity of the photodetectors (see Appendix A1.4).

2.4. Transition Radiation Detectors (TRD)

The TRDs will serve for particle tracking and for the identification of high energy electrons and positrons (γ > 2000) which are used to reconstruct J/ψ mesons. The technical task is to develop highly granular gaseous detectors which deliver signals within less than 100 ns. Such high-speed counters are needed for the inner part of the detector planes which cover forward emission angles. For example, at a distance of 4 m from the target, we expect at small angles particle rates of about 140 kHz/cm2 for 10 MHz minimum bias Au+Au collisions at 25 AGeV (see appendix A1.1). New concepts for fast gaseous detectors are under investigation (see Appendix A1.3). For the detectors covering large angles - corresponding to about 70% of the detector surface - ALICE type TRDs could be used if they turn out to be less cost intensive.

2.5. Resistive Plate Counters for Time-of-flight measurements (RPC)

The TOF stop detector of CBM will have an active area of about 150 m2 when located at a distance of about 10 m from the target. At small deflection angles the pad size is about 4 cm2 corresponding to an occupancy of below 5% for central Au+Au collisions at 25 AGeV. For 10 MHz minimum bias collisions the innermost part of the detector has to work at 25 kHz/cm2. The required time resolution should be well below 100 ps. The technical challenges for RPC development are the rate capability, long term stability and realisation of large arrays with overall excellent timing performance. Recent achievements in R&D are presented in the Appendix A1.5.

2.6. Electromagnetic calorimeter (ECAL)

The electromagnetic calorimeter will be used to measure direct photons, neutral mesons decaying into photons, electrons and muons. Simulations and R&D have been started based on the shashlik type of detector modules as used in HERA-B, PHENIX and LHCb. Particular emphasis is put on a good energy resolution and a high pion suppression factor (see Appendix A1.6).

2.7. Diamond pixel detector for TOF start signal

A diamond pixel (or micro-strip) detector provides the start signal for the TOF measurement. It will count directly the beam particles up to intensities of 109 ions/s.

13

3. Trigger, Data Acquisition and Event Reconstruction

3.1. Trigger and DAQ design considerations and developments

The key requirement driving the trigger and DAQ architecture of the CBM experiment is the efficient detection of rare probes like open charm, J/Ψ mesons, or low-mass dilepton pairs in the high-multiplicity environment of a heavy-ion collision. A trigger for these signals cannot be based on a fast preselection, determinable from a small subset of the detector information, but requires the evaluation of complex signatures, like track reconstruction or vertex determination for each interaction.

This implies an overall system design in which all detector channels act as self-triggered entities, autonomously detect signals and pre-process them to extract the physically relevant information. The data related to a particle hit are marked by a precise time stamp and transmitted to a buffer pool. The trigger processing occurs in compute nodes which access the buffer pool via a high-bandwidth network fabric.

This concept provides a high degree of flexibility in the choice of trigger algorithms. It makes first level trigger conditions available which are outside the capabilities of the standard approach, like for example an open charm trigger based on displaced vertex search. In addition, all detector subsystems can contribute to the trigger decision on the same footing, which is of particular relevance for the implementation of a low-mass dilepton trigger. Equally important is that the essential performance limitation of the trigger system is throughput and not latency, resulting in a much better utilisation of the resources than in a conventional approach. This is especially true in the case of heavy-ion collisions with their strongly varying multiplicities. Last but not least, this concept can be implemented in a highly modular fashion. A small number of building blocks dedicated to specific functions like readout, buffering, transfer or computing can be combined in a flexible way to cope with varying demands.

The main elements of such an architecture are developed in the context of the FutureDAQ Joint Research Activity within the EU 6th Framework Integrated Infrastructure Initiative for Hadron Physics (I3HP) in a joint effort of the CBM, PANDA and COMPASS experiments. The project leverages the extraordinary recent and projected progress in two key technologies:

• High-density programmable logic chips: FPGAs with a capacity well above 100k logic cells at commodity prices will allow to execute parts of complex trigger algorithms, like vertex reconstruction, in programmable logic.

• High-speed serial links: 10 GBit/s links over optical or copper interconnects with inexpensive serialisers in CMOS will allow to implement network fabrics with the required throughput in the Tbyte/s range.

The development is focused on three areas:

• Hardware building blocks:

• Frontend modules capable to perform self-triggered hit detection, pre-processing and feature extraction. The large channel density in many CBM sub-systems will require ASIC developments.

• Time distribution system. Via a passive optical distribution system a precise clock and absolute time are provided to all frontend modules. A global jitter below 25 ps is mandatory to meet the demands of precision time-of-flight measurements.

14

• Compute nodes. They perform the extraction of physical signatures and generate the trigger decisions. Several types with a different mix of programmable logic, processors, memory, and networking resources will be investigated. We will explore arrangements with directly interlinked compute nodes as well as concepts based on standard PC's with attached FPGA processors.

• Network fabric. The development of an embeddable optical transceiver and a cascadable cross-bar switch will allow a very cost effective integration of communication functions in frontend modules as well as buffer and compute nodes.

• Algorithms. The final trigger system will be a parallel computing system build from layers of programmable logic and processors. This requires to split the algorithms into local, regional and global parts and to map them accordingly to their computational and data access characteristics onto the most efficient compute architecture. The current research includes the development of highly pipelined and parallelised algorithms that execute on FPGA networks as well as algorithms for processors.

Figure 3.1: Schematic layout of the DAQ system A possible implementation of this architecture is sketched in Figure 3.1. The frontend modules are all synchronised by the time distribution system, process the detector signals, and ship the parameters of detected hits, with time stamps attached, to concentrators. This layer bundles the data from a group of frontends, performs, if applicable, some local level processing, like cluster finding, to reduce the data volume, and transmits the data, still at the full interaction rate, to buffer modules. This stage not only stores data for retrieval by the trigger processors, but will also perform some processing, like tracklet finding or event tagging, and provides specific data subsets and representations optimised for efficient trigger processing. The layer is therefore labeled 'Active Buffers'. The first level trigger decision, to be performed at the full design interaction rate of up to 10 MHz, will be done in an event-

15

parallel processor farm labeled 'Level 1 processors'. A key element of the architecture is that these processors will be able to selectively access over the 'Level 1 Network' all detector information, giving full flexibility for the trigger development and avoiding hard-wired connectivity restrictions. It is prudent to split up the trigger decision in several levels, where in each level the event rate is reduced, thus allowing more complex algorithms involving a larger amount of data in the subsequent level. The different nature of algorithms and data access patterns may lead to the use of different types of processors and networks in the different trigger levels, as indicated in Figure 3.1. The most cost effective layout for CBM and the choice of compute and network technologies used in the various layers will be done at the end of the 3 year R&D phase of the FutureDAQ project. 3.2. Event reconstruction

The process of track reconstruction can be broken down into several largely independent

steps. Some of these steps can (or even must) be performed in real time by dedicated processors, others are better for running on CPU or at the off-line stage of the data analysis. In the first step, the data of each detector are converted to spatial coordinates or space points, depending on type of a detector. Local preprocessing makes compression of data into clusters and finds tracklets. The second step is a local pattern recognition within sub-detectors. The third step consists of global track finding and fitting making full event reconstruction in the setup. The task is to reconstruct up to 700 tracks (central Au+Au collisions at 25 AGeV) from the particle hits measured in the various detector stations of the CBM experiment. An example of track reconstruction in the Silicon Tracking System is shown in figure 3.2.

Figure 3.2: Reconstructed tracks in the Silicon Tracking System for a central Au+Au collision at 25 AGeV

Figure 3.3: Track reconstruction efficiency as function of momentum in GeV/c

Global methods of track recognition use a description of a track by a set of its parameters.

Once the track model and detector measurement model are given, all hits in the detector can be projected into the track parameter space creating a complex density distribution with many local maxima. In this case the track recognition becomes a search for the local maxima corresponding to tracks.

Global methods are essentially maximum likelihood algorithms of parameter estimation. This property makes them, in principle, the most robust techniques for pattern recognition for simple geometries which allow parametrisation, for instance, of straight lines, parabolas, or circles. We have implemented one of the well-known global methods, the Hough Transform, in order to investigate its hardware (e.g. FPGA) applicability in a level-1 trigger for CBM. Alternatively, the Conformal Mapping Method can be used to transform a complicated

16

particle trajectory into a simple straight line or a parabola. We are presently also exploring this option.

Semi-global methods try to enhance the efficiency of global methods by local formation of

space points or short track segments in adjacent detector planes. The algorithm employed here is a Cellular Automaton method. Being essentially local and parallel, cellular automata avoid exhaustive combinatorial searches, even when implemented on conventional computers. Since cellular automata operate with highly structured information, the amount of data to be processed in the course of the track search is significantly reduced. Further reduction of information to be processed is achieved by a smart definition of the segment neighbourhood. Usually, cellular automata employ a very simple track model which leads to utmost computational simplicity and a fast algorithm. As a first result, the efficiency of track reconstruction for particles detected in at least three stations of the CBM Silicon Tracking System is presented in figure 3.3. Tracks of high-momentum particles are reconstructed very well with an efficiency of about 98%, while multiple scattering in the detector materials leads to a lower reconstruction efficiency for slow particles.

A parabolic approximation is used in the fitting procedure, which gives a momentum resolution of 0.6%. Another fitting algorithm based on the Kalman Filter is under development.

Different methods can be used to reconstruct rings in RICH detectors. The development of the reconstruction algorithms includes the following steps:

• Development of numerical algorithms for calculation of 3D components of the magnetic field of the superconducting dipole magnet.

• Implementation of the Kalman filter, Hough transform and cellular automaton methods for the track recognition problem in non-homogeneous magnetic field.

• Development of a robust approach of the Kalman filter for tracking at high track multiplicities.

• Investigation of quality of the Kalman filter for track parameter estimation in non-homogeneous magnetic field.

• Investigation of quality of the Kalman filter reconstruction of low momentum particles with significant multiple scattering.

• Development and comparison of three Cherenkov ring recognition algorithms: conventional one, based on robust fitting of Cherenkov rings to previously calculated hits; novel, based on direct processing of the raw RICH data. Both algorithms suppose to apply the prior knowledge about approximate position of Cherenkov ring centers and particle momenta obtained from out-of RICH tracking detectors. A third, standalone, algorithm of ring recognition will be also developed.

• Development of fast algorithms of primary and secondary vertices reconstruction. As an example, figure 3.4 depicts the result of a standalone ring search program.

Figure 3.4: Rings reconstructed by a standalone ring search program

17

4. Physics performance

4.1. Detector simulation software package

Our first GEANT4-based detector simulation package for CBM (G4CBM [1]) uses a simplified geometry and includes the CBM detector components STS, RICH, TRD, TOF and ECAL. The simulation output is written in ROOT format and can thus be analysed with this tool. The geometry is provided from an ASCII database in order to simplify the definition of subdetectors or passive volumes. Magnetic field maps as calculated by the TOSCA code can be used as input. The schematic CBM detector setup as used in the simulation is shown in figure 4.1.

Figure 4.1: The CBM detector setup as implemented in GEANT4 To reduce the dependence of our physics simulations on a specific simulation tool, we are currently investigating the possibility to use Virtual Monte Carlo (VMC), a ROOT-based interface developed by the ALICE collaboration that allows to run different Monte Carlo engines (GEANT3, GEANT4, FLUKA) without changing the application specific codes such as geometry definition, detector response simulation, or input and output format. Currently, a VMC application for CBM is being developed in parallel to the pure GEANT4 based application. This will permit to easily compare results obtained with e. g. GEANT3 and GEANT4, which may differ due to the different implementation of physics processes in these tools. However, major modifications had to be introduced to the VMC-GEANT4 branch to meet the CBM requirements. The medium term goal is to replace the pure GEANT4 based application with the VMC one.

As event generator we employ the UrQMD program [2]. In addition, the PLUTO package [3] is used for the generation of dilepton sources. Unless stated otherwise, the feasibility studies presented in the following sections refer to central Au+Au collisions at 25 AGeV processed through the experimental setup by the G4CBM tool.

18

4.2. Hadron identification by TOF

The determination of the particle mass is based on the measurement of the time of flight, the particle momentum and the particle track length. In the following we concentrate on the separation of pions and kaons, since protons are easier discriminated by their larger mass. The kaon identification efficiency and the amount of contamination by pions depends on the width and position of the window in the (squared) mass spectrum (see figure 4.2).

Fig: 4.2 Yields of pions (B= background) and kaons (S= signal) as function of the squared mass. The width and the position of the analysis window define the kaon identification efficiency and the purity: p = nS/(nS+nB)

The determination of the momentum dependence of the kaon identification efficiency is based on the following assumptions:

1. Particle yields from UrQMD for Au+Au at 25 AGeV: 297 π+, 325 π¯, 32 K+, 14 K¯ 2. time resolution of 60 ps and 80 ps 3. momentum resolution of σp/p = 6.4·10-3 4. resolution of 4 mm in track length determination 5. purity of 50 % (e. g. for the identification of secondaries from D meson decay)

Figure 4.3a shows the K¯ identification efficiency as function of laboratory momentum for two different assumptions of the track length (TOF distance 10 m and 15 m, respectively). The result is normalized to the number of geometrically accepted kaons. While the position at 15 m improves the mass resolution and therefore increases the efficiency for high momentum kaons, there is on the other hand a substantial loss of (mainly slow) kaons due to their decay in flight. A similar improvement in momentum coverage would be obtained if the time resolution could be decreased to 60 ps (see figure 4.3b). Such a time resolution is very challenging but would permit to reduce the distance from the target to the TOF wall and hence the detector area and the construction costs, as well as the decay losses at small kaon momenta.

Figure 4.3a: K- identification efficiency as function of laboratory momentum at 10 m and 15 m distance for σtof = 80 ps

Figure 4.3b: K- identification efficiency as function of laboratory momentum at 10 m for σtof = 60 ps and 80 ps

p (GeV/c) p (GeV/c)

19

Multi-strange baryons

The Silicon tracker station will permit to identify (multi)-strange hyperons via their decay topology. Feasibility studies are in progress to optimise the efficiency of the system. 4.3. Low-mass vector meson identification

The challenge in the measurement of low-mass vector mesons (ρ, ω, ϕ) by their decay into an electron-positron pair is the suppression of the combinatorial background. This background is due to electrons and positrons from other physical sources, among which the π0 Dalitz decay and the conversion of γ from π0 decays in the target and detector material are the most prominent ones. In addition, charged pions misidentified as electrons contribute to the background.

Due to the small dielectron branching ratios of the vector meson decays (≈ 10-5 - 10-4), a sophisticated cutting strategy has to be developed for the background suppression. Electrons from Dalitz decays can be effectively reduced by a cut on the transverse momentum since their pt spectrum is significantly softer than that of signal electrons. Low-mass (conversion) pairs are rejected by a cut on the pair opening angle. A preliminary analysis, using a parameterized detector acceptance and ideal efficiency and taking into account most physical dielectron sources together with an estimate for conversion electrons based on a detector simulation (see table 4.1) was used to develop a sophisticated cutting strategy. It yields a signal-to-background ratio of 0.5 – 1, strongly depending on the amount of conversion electrons.

source BR pairs / event π0→e+e-γ 1.198·10-2 7 η0→e+e-γ 5.0·10-3 0.49 η’→e+e-γ 3.9·10-4 0.02 ω→e+e-π0 5.9·10-4 0.048 φ→e+e-η 1.3·10-4 1.2·10-3 η’→e+e-ρ 2.0·10-3 0.11 ρ→e+e- 4.44·10-5 3.89·10-3 ω→e+e- 7.07·10-5 5.79·10-3 φ→e+e- 3.1·10-4 2.89·10-3

DY→e+e- 1.46·10-4 J/Ψ→e+e- 5.26·10-4 π+- misid. 0.25

Table 4.1: Average number of e+e- pairs produced via the decay of various particles in the parameterized detector acceptance

In a second step, the detailed detector setup was taken into account by using the simulation tool G4CBM and thus accounting more realistically for γ conversion in the detector. We used the UrQMD event generator to calculate hadron production rates and the PLUTO generator for the dilepton signal sources. The number of generated UrQMD events and the number of generated PLUTO events are different and a dedicated event merging procedure is applied. The simulation and the analysis of the results are based on the following simplifying assumptions:

• 100% efficiency of electron identification,

20

• perfect track and momentum determination, • no magnetic field, • gold target thickness of 100 µm, • vacuum beam pipe made of carbon fibre with a wall thickness of 0.5 mm, • seven silicon tracking stations with a thickness of 100 µm each, placed at the

distances of 5, 10, 20, 30, 40, 80 and 100 cm downstream from the target.

The calculations were performed for central Au + Au collisions at 25 AGeV. For the electron/positron background we took into account Dalitz decays of neutral π, η and ω, and γ conversion in the material of the target, beam pipe, magnets and Silicon Tracking System.

In order to suppress the combinatorial background, the following cuts have been applied to single tracks or track pairs:

• the e+ or e- is rejected if its transverse momentum is smaller than 0.1 GeV/c, • the e+ or e- is rejected if its track does not point to the interaction vertex, • the e+e- pair is rejected if the angle between the e+ and e- tracks is smaller than 10

degrees.

Figure 4.4 presents the e+ e- invariant-mass spectrum resulting from 108 central Au+Au events of 25 AGeV after these cuts. (black dashed line). The combinatorial background, obtained by the so called same-event procedure (solid red line) is clearly dominating the spectrum. The background-subtracted spectrum, shown by the histogram, satisfactorily reproduces the signal input to the simulation as indicated by the blue dashed-dotted line, thus confirming the background subtraction procedure. From this analysis, we derive a signal-to-background ratio of about 0.16 for ρ+ω and about 0.19 for ϕ. In the next step, tracking in the magnetic field and particle identification will be taken into account.

Figure 4.4: Electron pair mass spectra from a central Au+Au collision at 25 AGeV: all pairs (dashed black line), reconstructed combinatorial background (solid red line), polynomial fit to the background (green line), subtracted spectrum (full dots), vector meson signal input (blue dashed-dotted line).

21

4.4. D meson detection

One of the major tasks of the CBM experiment will be the measurement of open charm. D mesons will be detected via their weak decay into charged pions and kaons. The difficulty of this measurement lies in the very low multiplicity of D mesons (~10-3 per central Au+Au event at 25 AGeV as predicted by the HSD model [4]) within an environment of ~900 charged hadrons produced in such a collision. Evidently, the combinatorial background stemming from directly produced particles has to be suppressed by many orders of magnitude.

To study the feasibility of a D0 measurement in CBM, we have simulated the signal as well as the background composed of combinations of directly produced π+ and K¯ mesons. Both were processed through the GEANT4 based detector simulation tool. As a first step, we assumed perfect particle identification and tracking. No magnetic field was taken into account, so the tracks were fitted by straight lines.

For the background subtraction, we considered cuts in the following variables of the particle pair:

• The angle of the positive particle with respect to the pair momentum direction, calculated in the pair rest frame (decay angle). For an isotropic decay, the cosine of this variable should be evenly distributed, while it sharply peaks at small angles for background pairs.

• The product of the impact parameters of both tracks, defined as the distance of the track extrapolation to the interaction vertex in the target plane. While the tracks of directly produced particles should point back to the target, this is not the case for secondaries from D0 decays.

• The angle between the displacement vector of the secondary vertex and the pair momentum. In the case of a real D0 decay, both should be collinear.

• The z position of the secondary vertex, defined as the point of closest approach of two tracks. This is the most crucial variable to discriminate background from signal pairs. The z coordinate corresponds to the beam axis.

Figure 4.5: Distribution of the z coordinate of the two-track vertex of signal pairs (left) and background pairs (right)

The mean lifetime of the D0 mesons is cτ = 124 µm, which means that the secondary vertex needs to be determined with a resolution of less than ~50 µm. This can be achieved with the two first stations of the tracking system, which are placed 5 cm and 10 cm from the interaction point, respectively. The secondary vertex resolution depends on the material in the first station, causing multiple scattering, and the intrinsic single-hit position resolution of

22

the tracking detectors. Figure 4.5 shows the distributions in the secondary vertex position for signal and background pairs, assuming 100 µm thickness of the first station and a position resolution of 5 µm, which are typical parameters for MAPS detectors. For these distributions, the optimal cut would be at vz = 0.3 mm. However, one should note that weak decays of hyperons, which also produce an asymmetric vz-distribution, are not yet taken into account in the simulation.

We derive a background suppression of 2·105 after all cuts in the signal mass region; the signal efficiency is about 50%. The resulting signal-to-background ratio is about 2. The D0 detection rate would be 14,000 per hour at 1 MHz interaction rate for central events. However, this number largely depends on the production yield of charmed mesons, which can easily differ by an order of magnitude depending on which model is used for the prediction.

Since the ratio of π¯/K¯ is about 30, the signal-to-background ratio would decrease by approximately this factor if no particle identification were available. Thus, the identification possibility by TOF has been studied. As shown in figure 4.6, a large part of the secondary kaons can be identified up to 4 GeV laboratory momentum. For larger momenta other methods have to be used for pion suppression in the combinatorial background.

Figure 4.6: Momentum distribution for kaons from D0 decays. The dotted curve shows all kaons, the dashed curve those accepted by the tracking system, the full curve kaons accepted by the TOF system and the shaded area those identified by time-of-flight.

A similar study has been performed for charged D mesons. The main cut variable in this case is the impact parameter of the track extrapolation in the target plane. The longer mean lifetime (cτ = 317µm) and the decay charge topology allow the detection even without particle identification. We derive a signal-to-background ratio of roughly 3.

4.5. J/ψ detection

The feasibility study of charmonium measurements with the CBM detector via the detection of electron-positron pairs has been performed for Au+Au collisions at 25 AGeV. The event generator UrQMD was used to calculate the multiplicities and transverse momentum distributions of charged pions. The event generator PLUTO was used to calculate the decay kinematics of electrons and positrons originated from J/ψ mesons and ρ mesons and from the Dalitz decays of π0 and η mesons. In addition, the π0 decay into γγ was taken into account. The interaction of the particles with the materials of target, Silicon Tracking Station, beam pipe etc. was calculated with the detector simulation tool G4CBM, including the conversion of γ from π0 decays into e+e- pairs.

p (GeV/c)

23

The single electron yields and momentum distributions were calculated taking into account the dileptonic decay of ρ mesons and open charm, the Dalitz decay of π0 and η mesons, conversion of γ in the target (Au 0.3 mm thickness), and misidentified charged pions. The pion suppression factor was assumed to be 10-4 (1% for both RICH and TRD). The resulting transverse momentum distributions are shown in figure 4.7. These spectra were obtained from 10000 UrQMD events for the hadronic contributions, and from 5000 PLUTO events to get the ρ, open charm and J/ψ contributions. The particle multiplicities shown in figure 4.7 were normalised to one UrQMD event.

The combinatorial background in the invariant-mass spectrum was calculated from the single electron momentum distributions given from 15000 UrQMD events and is shown in figure 4.8. The dominating contribution is due to γ conversion in the 0.3 mm Au target. This contribution could be reduced by using a segmented target. The J/Ψ mass spectrum was calculated assuming a momentum resolution of 1 %. A high statistics GEANT4 simulation is in progress.

Figure 4.7: Number of measured single electrons per event as function of transverse momentum in GeV/c for Au+Au at 25 AGeV. J/ψ (red), ρ (pink), misidentified charged pions (green), π0 Dalitz (violet), η Dalitz (blue), conversion (black) Yields are per event. The temperature of π and η in PLUTO is assumed to be 130 MeV and for J/ψ - 170MeV

Figure 4.8: Dilepton invariant mass spectra without cuts (black histogram) and after applying the cut on pt > 1 GeV/c and on pair opening angle θ > 10° (cyan histogram).

24

The combinatorial background can be suppressed by cutting on the single electron transverse momentum and on the pair opening angle. In this study, only electrons and positrons with pT larger than 1 GeV/c and with a pair opening angle larger than θ = 100 were accepted. The resulting invariant mass distribution for electron/positron pairs with single electron/positron transverse momenta larger than 1 GeV/c and with opening angles of θ > 100 is shown in figure 4.8 (cyan histogram). A more realistic simulation with increased statistics and taking into account tracking and particle identification is in progress. In addition to the measurements of electrons we will study the option of detecting muons in order to identify charmonium. References [1] http://www-linux.gsi.de/~gsisim/cbm.html [2] S. A. Bass et al., Prog. Part. Nucl. Phys. 41 (1998) 225 [3] http://www-hades.gsi.de/computing/pluto/html/PlutoIndex.html [4] W. Cassing and E. Bratkovskaya, Phys. Rep. 308 (1999) 65

25

5. Experimental programme and beam time estimate

5.1. Beam requirements

The research programme to be performed with the CBM experiment requires the following accelerator performances:

1. intensities of heavy-ion beams (up to Uranium) of 109 particles per second up to the maximum beam energy of 35 AGeV for U and 45 AGeV for nuclei with Z = 0.5 A,

2. proton beams with intensities of 1010 protons per second up to an energy of 90 GeV,

3. beam spot at the target position not larger than 5 mm2,

4. beam profile without halo, i.e. the intensity outside a radius of 5 mm from the beam axis should be below 10-6 of the total beam intensity.

5.2. Experimental programme

The envisaged experimental programme includes the following topics:

• compressed baryonic matter studies: systematic investigation of nucleus-nucleus (A+A) collisions at beam energies ranging from 2 - 45 AGeV (Z/A = 0.5) and up to 35 AGeV for Z/A = 0.4;

• reference measurements on the production of charm, strangeness and low-mass vector mesons in proton-nucleus collisions up to energies of 90 GeV (nuclear matter at saturation density);

• measurement of elementary production cross sections of charm (charmonium, D Mesons), light vector mesons, multistrange hyperons, and exotica like pentaquarks etc. in proton-proton collisions at beam energies up to 90 GeV.

The Data Acquisition System (DAQ) of CBM will be able to record minimum bias events at a rate of 25 kHz. A high-speed trigger system will be used for the measurement of rare probes such as charmonium, and D mesons.

5.3. J/ψψψψ meson measurements

The J/ψ multiplicity is expected to be about 4⋅10-6 per minimum bias Au+Au collision at 25 AGeV [1]. Taking into account the branching ratio of 6% for the decay into a lepton pair and the efficiency (incl. acceptance) of 0.1 the detected J/ψ yield is about 2.4·10-8 per event. At a reaction rate of 5 MHz the number of recorded J/ψ mesons is about 104 per day.

For the J/ψ measurement the trigger will require an (identified) e+e- pair with a transverse momentum above 1 GeV/c for each lepton and an opening angle of the pair of larger than 10 degrees. The dominating background is due to gamma conversion. The probability of having an electron-positron pair (with lepton pT above 1 GeV/c ) from gamma conversion is about 10-5 per event. This means that a primary reaction rate of 5 MHz is reduced down to 50 Hz by the trigger requirements and, hence, can be handled easily by the DAQ system.

26

Table 5.1: Run time estimate for J/Ψ measurements in Au+Au collisions The total run time for the J/ψ measurements in the Au+Au system amounts to about 1 year including commissioning (see table 5.1). Another year is planned for an intermediate mass system (A=100) and for a light collision system such as C+C up to 45 AGeV beam energy. The integral run time for J/ψ measurements in A+A collisions is about 2 years.

One year of run time is needed for the J/ψ measurements in proton-nucleus collisions using three different nuclear targets and about 10 proton beam energies. Several months of run time is needed for the measurement of the J/ψ production excitation function in proton-proton collisions at beam energies close to the threshold.

5.4. D meson measurements

According to a prediction of a HSD transport calculation [1] the D0 meson multiplicity is about 2·10-4 per minimum bias Au+Au collision at 25 AGeV. Taking into account the branching ratio of about 4% for the decay into a kaon-pion pair and an efficiency (incl. acceptance and trigger) of 0.1 the detected D0 yield is 8·10-7 per event.

Our goal is to develop a D0 meson trigger based on the displaced vertex of the kaon and the pion. This requires high-resolution online tracking and particle identification. The aim is to reduce the primary reaction rate of 10 MHz to 25 kHz by the trigger (suppression factor of 400). Consequently, the recorded number of D0 mesons in Au+Au collisions at 25 AGeV is about 8 per second and about 7·105 per day. In addition we will measure charged D mesons via their decay into a kaon and two pions. These measurements can be performed in parallel to the J/ψ experiments. Moreover, one can study open charm correlations: the measured number of D+D- pairs is about 100 times lower than the number of single D+ or D- mesons (efficiency ≈ 0.1, branching ratio ≈ 0.1).

The D meson research programme includes also p+A and p+p collisions. These data are particularly important as a reference for the A+A data. No data on D meson production are available at proton energies below 200 GeV.

5.5. Low-mass vector meson measurements

The ρ-meson multiplicity is about 4 per minimum bias Au+Au collision at 25 AGeV. The branching ratio for the decay into an electron-positron pair is 4.5·10-5. Assuming a detection efficiency of 0.1 (incl. acceptance and trigger) the number of measured ρ-mesons is expected to be about 1.8·10-5 per event.

beam energy AGeV

J/ψ mult. produced min. bias

J/ψ yield detected per week

runtime weeks

10 4·10-8 7·102 10 15 4·10-7 7·103 10 20 2·10-6 3.5·104 10 25 4·10-6 7·104 5 30 1.2·10-5 2.1·105 2 35 2·10-5 3.5·105 1

27

The combinatorial background in the electron-positron invariant mass spectrum has two major sources: Dalitz decays of the π0 (and η)-mesons and γ conversion in the target and the detector materials. The gamma rays stem predominantly from π0 decays. Due to the large electron background there is a high probability to find Cherenkov rings in each event. Consequently, a trigger on double rings will not reduce considerable the primary reaction rate. Therefore, for the following beam time estimate we do not assume to use a trigger. With a primary minimum bias event rate of 25 kHz (2,2⋅109 events per day, corresponding roughly to the statistics in figure 4.4), we expect to record about 4·104 ρ mesons per day for Au+Au collisions at 25 AGeV.

The production of low-mass vector mesons will be systematically studied in A+A, p+A and p+p collisions for various A in the beam energy range from 8 to 45 AGeV (A+A) and up to 90 GeV for proton beams. Assuming a run time of 2 weeks for each A+A system with A = 200, 100, 12 and 6 energy steps the total run time is about 36 weeks beam on target. For the proton induced reactions we expect a total run time of about 20 weeks. The whole research program will require about one year beam on target.

5.6. Hyperons, antiprotons, fluctuations, hadron flow, exotica

No dedicated trigger is required for the measurement of (multi-strange) hyperons, of antiprotons, of fluctuations and of hadron flow. For example, the multiplicity Ω- baryons in Au+Au collisions is about 4·10-2. Assuming a detection efficiency of 1 % and a minimum bias event rate of 25 kHz we can collect about 10 Ω- per second. Billions of minimum bias events can be taken within one day. The data can be taken either in short dedicated runs or in parallel to other measurements with a down-scaled minimum bias trigger.

5.7 Medium energy research programme

The beam energy range between 2 and 8 AGeV can be studied with the HADES spectrometer. The research program includes the measurement low-mass-vector mesons (ρ,ω,ϕ) and hadrons in nucleus-nucleus, proton-nucleus and proton-proton collisions. The integral run time is expected to be about 2 years. References [1] W. Cassing and E. Bratkovskaya, Phys. Rep. 308 (1999) 65

28

6. Responsibilities, schedule

6.1 Responsibilities

Coordinator: P. Senger

Coordinators of working groups: Simulations: R. Holzmann Silicon Tracker: J. Stroth Hadron identification: N. Herrmann Electron identification: J. Wessels Trigger/DAQ: W. F. J. Müller

Contributions to this LOI:

Feasibility studies Simulation tools: M. Al-Turany, D. Bertini, V. Friese, K. Schwarz Hadron ID: N. Herrmann, D. Kresan, K. Wisniewski Low mass vector mesons: H. Appelshäuser, A. Baldin, E. Baldina, J. Cibor, R. Karabowicz,

P. Staszel, P. Szostak, Z. Majka J/Ψ mesons: F. Guber, E. Karpechev, A. Maevskaya D mesons: V. Friese, V. Petracek

Trigger/DAQ I. Kisel, W. F. J. Müller

Detector design and development Silicon Tracker: M. Deveaux, J. Stroth, M. Winter RICH: S. Sadovsky TRD: A. Andronic, V. Peshekhonov, J. Wessels, Y. Zanevsky RPC: P. Fonte, N. Herrmann ECAL: A. Golutvin, S. Kiselev, I. Korolko Magnet: M. Al-Turany, P. Akishin HADES@SIS100/300: D. Adamova, A. Kugler, P. Tlusty Editors: V. Friese, P. Senger 6.2 Funds CBM related R&D projects are supported by the following funds: EU FP6 Hadrons 2004-2006:

• Fast gas detectors: Spokesperson J. Wessels • New TOF detectors (RPC): Spokesperson N. Herrmann • High Speed trigger/DAQ: Spokesperson W.F.J. Müller • CBMnet: Spokesperson P. Senger

INTAS/GSI 2004-2005:

• Transition radiation detectors: Coordinator C. Garabatos • Resistive Plate detectors: Coordinator N. Herrmann • Straw tube tracker: Coordinator L. Naumann • Electromagnetic calorimeter: Coordinator J. Eschke

29

6.3 Working packages

Task

Institution

Feasibility study D-Meson identification GSI Darmstadt, Czech Acad. Science Rez,

Techn. Univ. Prague Feasibility study low-mass vector meson identification via dilepton pairs

Univ. Krakow, JINR-LHE Dubna

Feasibility study charmonium identification INR Moscow

Simulations hadron identification via TOF Heidelberg Univ., Kiev Univ., NIPNE Bucharest, INR Moscow, RBI Zagreb

Simulation tools GSI Darmstadt

Tracking KIP Univ. Heidelberg, Univ. Mannheim, JINR-LHE Dubna, JINR-LIT Dubna

Silicon Pixel Detector IReS Strasbourg, Frankfurt Univ., GSI Darmstadt, RBI Zagreb, Krakow Univ.,

Silicon Strip Detector SINP Moscow State Univ., CKBM St. Petersburg, KRI St. Petersburg

R&D on RPC TOF detector system with read-out electronics

LIP Coimbra, Univ. Santiago de Compostela, Univ. Heidelberg, GSI Darmstadt, NIPNE Bucharest, INR Moscow, FZR Rossendorf, IHEP Protvino, ITEP Moscow, Korea Univ. Seoul, RBI Zagreb, Univ. Krakow,Univ. Marburg

R&D on fast gaseous detectors for TRD

JINR-LHE Dubna, GSI Darmstadt, Univ. Münster, PNPI St. Petersburg, NIPNE Bucharest

R&D on straw tube tracker (TRD) JINR-LPP Dubna, FZR Rossendorf

R&D on Ring Imaging Cherenkov Detector (RICH)

IHEP Protvino, GSI Darmstadt, Pusan Nat. Univ., PNPI St. Petersburg

Design and construction of an electromagnetic calorimeter (ECAL)

ITEP Moscow, Univ. Krakow, Univ. Frankfurt

Diamond microstrip detector GSI, Univ. Mannheim

Trigger and Data Acquisition KIP Univ. Heidelberg, Univ. Mannheim, JINR LIT Dubna, GSI Darmstadt, Univ. Bergen, KFKI Budapest, Silesia Univ. Katowice, PNPI St. Petersburg, Univ. Warsaw

Design of a superconducting dipole magnet

JINR-LHE Dubna, GSI Darmstadt

Calculation of radiation doses Kiev Univ.

Modification of HADES for 8 AGeV Czech Acad. Science Rez

Delta electrons GSI Darmstadt

30

6.4 Time schedule Milestones: 1. Technical Proposal End of 2004 2. Technical Design Report End of 2007 Subproject 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012Si-Tracker RICH TRD TOF-RPC ECAL Trigger/DAQ Electronics Magnet Infrastructure

simulations, R&D, design Prototyping Construction Installation, test 6.5 Radiation Environment

The radiation dose distribution in the cave was calculated with the code FLUKA. The result is shown in figure 6.1. The shielding is sufficient to reduce the dose outside the cave to a level of 1 µSv/h.

Figure 6.1: Sideview of the cave. Radiation dose profile in a vertical plane along the beam line Due to the high beam intensities and the high reaction rates we have to consider radiation effects on detector materials and electronics. In particular, the silicon tracker and electronic components may be damaged by following effects:

• total Ionizing Dose (TID),

• displacement damages,

• single event effects (SEE) (such as Single Event Upset (SEU), Single Event Latchup (SEL), Single Event Burnout (SEB).

As a first step we have started to calculate particle fluxes and total ionizing doses. We use the code G4CBM and compare the result to calculations performed with the code FLUKA. The second step of the project is the implementation of classes to calculate the activation of materials.

31

Appendix 1: Detector studies

A1.1. Experimental conditions

The expected hit rates and hit densities have been calculated for Au+Au collisions at 25 AGeV using UrQMD. The results are shown in figure A1.1 for a detector at 1 m downstream of the target. In table A1.1 the values are given for different detector locations and emission angles.

Figure A1.1: Average hit density of charged particles for central Au+Au events at 25 AGeV (left) and hit rate of charged particles for minimum bias Au+Au events at 25 AGeV (right) assuming an interaction rate of 10 MHz. Both distributions are calculated at 1 m distance from the interaction point. The red lines show the density profile along the x axis, the blue curves along the y axis. The difference between the red and blue distributions is due to the magnetic dipole field.

TRD 1 dist. 4 m

emission angle [mrad]

rates [kHz/cm2]

area [m2] N [cm-2 ] cell size [cm2]

# of cells

50 140 0.2 6.3 10-2 0.8 2500 100 62 0.6 4.4 10-2 1.1 5450 200 25 2.3 1.1 10-2 4.5 5110 300 12.5 4.1 0.6 10-2 8.3 4940 400 6 6.3 0.3 10-2 10 6300 500 4 9.0 0.2 10-2 10 9000 Sum 22.5 33300

TRD 2 dist. 6 m

emission angle [mrad]

rates [kHz/cm2]

area [m2] N [cm-2 ] cell size [cm2]

# of cells

50 62 0.4 2.6 10-2 1.9 2100 100 27 1.3 2.0 10-2 2.5 5200 200 11 5.3 0.5 10-2 10 5300 300 5.5 9.3 0.25 10-2 10 9300 400 2.7 14.1 0.13 10-2 10 14100 500 1.8 20.2 0.09 10-2 10 20200 Sum 50.7 56200

32

Table A1.1: Hit rates, hit densities and detector areas at different detector locations and for different emission angles. The hit density is calculated for central Au+Au collisions at 25 AGeV. The rates are calculated for 10 MHz minimum bias Au+Au collisions at 25 AGeV. The cell size was determined by the requirement that the occupancy is below 5 %. The maximal cell size is 10 cm2 for the TRD and 20 cm2 for the RPC. The areas were calculated assuming an elliptical shape which takes into account the deflection of charged particles in horizontal direction, i.e. the horizontal dimension of the detectors is 1.5 times the vertical dimension. Rates and densities refer to charged particles for the TRD and the RPC and to the total number of particles (including neutrals) for the ECAL. A1.2. The Silicon Tracking System

The hit densities in the first and the last Silicon Tracking Station (5 cm and 100 cm downstream the target, respectively) are shown in Fig. A1.2. The figure presents the hit densities for primary and secondary charged particles and those for electrons emitted in central Au+Au collisions at 25 AGeV. The calculations were performed with the UrQMD event generator and the G4CBM code.

TRD 3 dist.8 m

emission angle [mrad]

rates [kHz/cm2]

area [m2] N [cm-2 ] cell size [cm2]

# of cells

50 35 0.8 1.6 10-2 3.1 2580 100 15 2.3 1.1 10-2 4.5 5110 200 6 9.3 0.28 10-2 10 9300 300 3 16.5 0.14 10-2 10 16500 400 1.5 25.0 0.08 10-2 10 25000 500 1 36.1 0.05 10-2 10 36100 Sum 90.0 94590

RPC dist 10 m

emission angle [mrad]

rates [kHz/cm2]

area [m2] N [cm-2 ] cell size [cm2]

# of cells

50 22 1.2 1.0 10-2 5 2400 100 10 3.6 0.7 10-2 7.1 5070 200 4 14.6 0.2 10-2 20 7300 300 2 25.7 0.09 10-2 20 12850 400 1 39.1 0.05 10-2 20 19550 500 0.6 56.4 0.03 10-2 20 28200 Sum 140.6 75370

ECAL dist 12 m

emission angle [mrad]

rates [kHz/cm2]

area [m2] N [cm-2 ]

50 15.5 1.7 1.3 10-2 100 6.8 5.1 0.72 10-2 200 2.8 21.1 0.29 10-2 300 1.4 37.0 0.11 10-2 400 0.7 56.4 0.05 10-2 Sum 121.3

33

The Silicon Tracking System consists of pixel detectors and strip detectors. The pixel detectors will cover the forward angle regions where the particle densities have values up to 2 hits per mm2 and event (see figure A1.2, left panel). The strip detectors will be located at larger angles or distances from the target where the hit densities are below 10-2 per mm2 and event (see figure A1.2, right panel). In the following, we describe our approach towards the development of a highly granular and ultra-thin pixel detector which is the necessary prerequisite for high-precision measurements of secondary vertices.

Figure A1.2: Hit density for central Au+Au reactions at 25 AGeV in the first plane of the Silicon Tracker Station (distance 5 cm from the target, left panel) and for the last plane (distance 100 cm, right panel) CMOS sensor application to vertex detector

Introductory remarks

The physics goals of the CBM experiment call for unprecedented flavour tagging performances, out of reach of existing technologies: Charge Coupled Devices (CCD) provide the necessary granularity and may be thinned down to the thickness wanted, but are too slow and radiation sensitive; Hybrid Pixel Sensors, on the contrary, are fast and radiation tolerant but suffer from modest granularity and significant material budget. The emergence of CMOS sensors for charged particle tracking offers new perspectives in high precision vertexing and may provide the required performances. Their development started only a few years ago and, though excellent detection efficiency and spatial resolution have been achieved, the questions of fast read-out and high radiation tolerance were only addressed very recently. Substantial R&D is therefore needed for solving these issues. The main features of CMOS sensors and of their demonstrated performances are summarised hereafter. The question of fast read-out is addressed next, followed by considerations on radiation tolerance. The issue of thinning is discussed at the end of the Appendix.

34

Main features and established performances

CMOS sensors are manufactured in standard CMOS technology, which offers low fabrication costs and fast turn-over for their development. The key element of this novel technology is the use of an n-well/p-epi diode to collect, through thermal diffusion, the charge generated by the impinging particle in the thin epitaxial layer underneath the read-out electronics [1]. An attractive peculiarity of the sensors is that they allow to fabricate System-on-Chips (SoC) by integrating signal processing micro-circuits (amplification, pedestal correction, digitisation, discrimination, etc.) on the detector substrate. Moreover, the latter may be thinned down to a few tens of microns since the active volume is less than 20 µm thick.

The ability of these sensors to provide charge particle tracking is now well established [2]: several prototypes, called MIMOSA (standing for Minimum Ionizing MOS Active pixel sensor) exploring different fabrication technologies and key parameters of the charge sensing, demonstrated that a detection efficiency of ≈ 99 % and a single point resolution of ≈ 2 µm could regularly be achieved, based on a pixel pitch of about 20 µm. It was shown that digitising the charge on a small number of ADC bits (e.g. 3 bits) would not degrade the resolution beyond ≈2.5 – 3 µm. The double hit resolution was also studied, and found to be ≈ 30 µm. Most of the R&D was performed with small (few mm2 wide) prototypes made of a few thousand pixels. A reticle size (i.e. ≈ 3.5 cm2) prototype, composed of ≈ 1 million pixel, was also fabricated in a batch of 6 inch wafers, as shown on figure A1.3.

The results of its tests at the CERN-SPS confirmed the performances obtained with the small structures: a 99 % detection efficiency and a ≈ 2 µm single point resolution were observed. This generation of sensors was not foreseen to allow very fast data treatment; it is well suited to running conditions requiring frame read-out times not much shorter than ≈1 millisecond. It fits, for instance, well the requirements of the vertex detector upgrades foreseen in the RHIC experiments in order to take advantage of the luminosity increase and beam pipe radius reduction planned in the coming couple of years. The use of CMOS sensors in the CBM experiment is much more demanding and imposes to shorten the read-out time by two orders of magnitude, a goal which calls for a vigorous R&D programme.

Figure A1.3: Wafer of reticle size sensors (left) and zoom on individual chips (right) The radiation tolerance of the sensors is still under study. At present, only very preliminary conclusions can be drawn. The effect of bulk damage [3] was investigated by exposing small prototypes to fluences of up to 1013 neq·cm-2. A decrease of the detection efficiency was observed, which started to be significant near 1012neq·cm-2. As far as ionising radiation damages are concerned, it is unlikely that any of the present prototypes can afford radiation doses exceeding 1 MRad. These results are illustrative of the natural radiation tolerance of the technology, but do not establish their ultimate resistance. A substantial R&D effort is

35

needed to reach it. The question of thinning was not yet much addressed, but the reticle size sensors mentioned above were actually thinned down to 120 µm before being tested, without any noticeable side effect. Thinning to 50 µm or less still needs to be demonstrated, but is not expected to be particularly difficult.

Development of fast sensors

The guide line leading to fast sensors consists in splitting them in numerous sub-areas treated in parallel. This approach translates into very large data flow, of the order a Tbit/s per sensor. The achievement of a high read-out speed therefore strongly depends on the possibility to implement on-chip hit recognition and data sparsification. On-chip processing is complicated by the smallness of the signal amplitude (in the range of millivolts), which is not far from natural dispersions occurring in CMOS processes. Another handicap comes from the constraint of the technology: N-wells are bound to serve as charge collecting diodes, a feature which hampers the use of p-mos transistors to treat the signal charge inside the sensor sensitive area. As a consequence, signal treatment inside pixels is limited to functionalities achievable with n-mos transistors, those relying on p-mos transistors being performed at the sensor edge, where no charge collection takes place. Since the charge collection time is dictated by thermal diffusion, and amounts to a few tens of nanoseconds, the ultimate read-out time achievable lies in the range of a few microseconds. The goal of the R&D is to reach this ultimate value.

The first prototype exploring a fast read-out architecture was fabricated in 2002 [4]. Each of its pixels is equipped with charge amplification micro-circuits and allows for correlated double sampling operation in order to subtract the pedestal associated to the average leakage current integrated by each sensing device. The chip is organised in columns grouping 128 pixels; its readout time is close to 25 µs. A comparator is integrated at the end of each column for discrimination purposes. Tests of the prototype showed very good noise performances at the single pixel level (i.e. ≈ 15 e- ENC during the read-out phase) as well as an acceptable dispersion of the discrimination thresholds. However, substantial additional noise was observed, presumably due to cross-talk between the digital and analogue circuits inside the pixels and to the dispersion of the pixel characteristics inside each column. Two new prototypes were designed and fabricated recently in order to improve the sensor performances. They will be tested in 2004.

The R&D in the coming years will concentrate on noise reduction and improved amplification (in order to reduce the influence of noise on the read-out chain). Various types of charge collecting devices will also be explored. A further step will consist in integrating an ADC circuit at the column ends. Next, data sparsification will be implemented. Finally, the integrated architecture allowing to store and extract the cluster information will be developed. All these functionalities will first be tested on small prototypes. Since they require a large number of transistors (i.e. several tens of millions) integrated on each chip, the fabrication yield will be an issue. Its assessment requires fabricating medium or reticle size sensors as well. The choice between different possible data treatment architectures will be guided by the aim of keeping the power dissipation well below 1 W/cm2.

The optimisation of the sensor performances will also depend on features specific to each fabrication process available. The latter need therefore to be explored continuously. For instance, a new fabrication technology, relying on a lightly doped substrate but exhibiting no epitaxial layer, was investigated with two prototypes. Their tests demonstrated an outstanding detection efficiency of up to 99.9 %, and a single point resolution of about 2.5 µm [5]. A major advantage of this technology is that the signal charge is large because several tens of µm depth contribute to its generation. This technology may thus be considered as very promising for chips hosting complicated read-out architectures producing sizeable electronic noise.

36

Radiation tolerance

Since the present knowledge of the sensor radiation tolerance is still very limited and needs to be better assessed, the two coming years will largely be devoted to the necessary measurements. These encompass the estimate of the tolerance itself as well as of the electrical parameters governing the reaction of the sensors to intense radiation, i.e. I-V, C-V and, more generally, reverse engineering measurements within the limits allowed by the chip manufacturers. The chips will be exposed to 10 keV X-Ray sources at CERN or in Karlsruhe, or to a neutron beam of 1 MeV average energy in Dubna. Possibilities exist also to irradiate the chips with an electron source.

The observed characteristics will be compared to, or used in, the 3-dimensionnal simulation programme ISE-TCAD, in order to investigate the mechanisms underlying the chip sensitivity. The goal is to understand sufficiently the dominating mechanisms at the origin of the chip performance loss in order to adapt the sensor design to enhanced radiation tolerance. The sensitivity of the sensors may strongly depend on the fabrication process. Each of them needs therefore to be carefully evaluated in terms of tolerance to bulk damage and ionising radiation effects. Clearly, priority will be given to manufacturing processes providing a feature size ≤ 0.35 µm. Transistors may be enclosed in order to protect them against polarity inversion induced by ionising radiation. However, this increases their size substantially, a feature which may lead to hampering micro-circuit surfaces. Studies will therefore be pursued in order to spot the transistors most exposed and vital which need to be enclosed.

The R&D effort will also touch the issue of procedures allowing to recover (at least partially) from damages due to radiation. Various treatments will be considered, which are mainly based on temperature and time considerations. The role of temperature needs actually to be investigated carefully, and the behaviour of the sensors will be studied as a function of operating temperature. Thinning

Thinning 6 inch wafers down to about 50 µm is not expected to be a big issue, as such possibilities seem quite widespread in industry nowadays. Thinning tests with wafers of the reticle size sensors mentioned earlier will be performed in 2004. Thinning below 50 µm may be considered. The ability of industry to achieve it with a satisfactory yield needs to be assessed. Moreover, internal mechanical stress may occur, translating into wrinkling of the chip. The answer to this side effect may consist in gluing the chip on a thin carbon foam or beryllium support, which would ensure the necessary rigidity. A dedicated research effort is needed to investigate such possibilities. It may profit from similar studies performed for the RHIC upgrades and for the Linear Collider project.

Framework of the R&D

CMOS sensors are being developed in Strasbourg since 1999 in perspective of various applications, which range from vertex detectors for subatomic physics (e.g. STAR upgrade, Linear Collider) to bio-medical imaging (e.g. beam monitoring for oncotherapy, dosimeters for brachyotherapy) and operational dosimetry (e.g. control of ambient radon and neutron radiation levels in nuclear plants). Several application domains call for SoCs providing fast read-out speed, high radiation tolerance, minimal material budget and low power dissipation. Developments for the CBM experiment will thus benefit from the synergy with the R&D aiming for other applications, in terms of fabrication process exploration, development of fast signal processing architectures, radiation tolerance investigations and improvements, etc. More information on the activities and achievements of the Strasbourg research team is available in [6].

37

A1.3. Transition Radiation Detector development

Requirements

Within CBM the Transition Radiation Detector (TRD) will be used for electron identification and discrimination against pions. The TRD is a gaseous detector and as such ultimately limited in its performance by the space charge built up by the positive ions in each readout cell. Owing to the fixed target geometry the angular distribution of particles is very much peaked at small angles as can be seen from tables A1.1, hence a segmentation with very different readout cell sizes is required.

In order to achieve the necessary pion rejection it is currently anticipated to build nine layers of TRDs. There will be three planes of detectors at distances of 4 m, 6 m, and 8 m from the target. The total area covered by TRDs is about 500 m2. The size of the readout cells has been adapted to achieve an occupancy of maximal 5 % and, on the other hand, is limited to 10 cm2 in order to obtain a good tracking performance (see table A1.1). This leads to about 550,000 channels in total. All of the detector technologies discussed below are capable of achieving a position resolution of order 200 µm. While here the proposed granularity is chosen to be roughly constant, the anticipated event rate of 107 minimum bias collisions per second leads to rates of up to 175 kHz for individual readout cells in central collisions. For peripheral collisions this will increase for the smallest angles. Therefore, the goal is to find detector designs capable of handling rates of up to 500 kHz. At the same time a pion rejection of several hundred at an electron efficiency of 90 % should be achieved.

Anticipated design options

The individual detector layers will be interleaved with fiber radiators. While the choice of the radiator is rather straight forward, it is also clear that a single technology for the readout chambers is not able to cope with the requirements outlined above. In the following we discuss the different options that allow to cope with the anticipated counting rates.

ALICE-type TRDs

In the region of lowest occupancy, it is conceivable to use ALICE-type TRDs which consist of a radiator, a small drift chamber, and a pad-plane with the full electronic readout chain [7-9]. The performance of these detectors in their final configuration is shown in figure A1.4.

Recently, it has been shown that the pion efficiency can be significantly improved by employing neural networks [10]. However, this type of detector will be limited to rates of at most a few kHz per readout cell. For CBM it is clear that, the design needs to be thinner and consequently the number of layers needs to be of order nine. Pion efficiencies as a function of number of layers is shown in figure A1.5 for equivalent fiber radiator thicknesses of 1, 2 and 3 cm. The total radiation length will be of order 0.2·X0 for the nine layers.

TRDs based on multiple ionisation stages

In order to overcome some of the limitations mentioned above it is proposed to study the performance of so-called micro-pattern detectors such as GEMs (Gas Electron Multipliers [11]), Micromegas (mesh-based gas detectors [12]), and others [13] using multiple stages for the amplification and working with very small gaps between these stages. To this end both an INTAS proposal [14] has been granted as well as a Joint Research Activity (JRA4 [15]) within the 6th Framework Program of the EU has been accepted. The anticipated designs allow to reduce ion drift times to order of 100 - 200 ns. The small readout structures will require an even higher level of integration as already foreseen for the ALICE-type detectors.

38

Figure A1.4: Pion efficiency as a function of momentum for different radiator sandwich types and a pure fiber radiator (open circles). In addition, the top panel also shows the separation without radiator (open crosses). The pion efficiency was extrapolated to six layers based on bi-dimensional likelihood distributions on total charge and position obtained in a single chamber.

Figure A1.5: Simulations of the pion efficiency as a function of the number of detector layers. The different curves refer to equivalent fiber radiator thicknesses of 1, 2 and 3 cm respectively (Nf =80, 160, 240).The thicknesses of the foils and gaps are d1and d2, respectively. The values of the regular radiator parameterization in simulations were tuned to reproduce measurements with fiber radiators. The horizontal line marks the required value of 1 % pion efficiency.

TRT-like designs

A typical representative of transition radiation detectors for very high rate applications is the ATLAS Transition Radiation Tracker (TRT) [16,17]. It has been designed to cope with counting rates of up to about 10 MHz, well above the requirements for this experiment. The major drawback of such a system is the relatively large size of the readout cells. A straw tube

39

of 40 cm length and 0.4 cm in diameter can be operated at a rate of 18 MHz corresponding to 1.1 MHz/cm2. For large systems, the position resolution is better than 150 µm. The radiation length of one straw plane is about 0.2 % (without gas). Test performed with 40 layers (ATLAS TRT) found a pion suppression factor of up to 300 for pions of 20 GeV/c. The hit efficiency of a straw tube is 85 – 95 %. Further simulations and R&D will concentrate on the reduction of the straw tube size and on the improvement of the efficiency. This activity is supported by an INTAS grant [18]. In the CBM environment it is conceivable that straws could be operated with multiple hits. However, in that case the readout then has to be via pads on the outside of the detector that pick up part of the induced signal.

Electronics

At event rates of 10 MHz it is obvious that the on-detector electronics needs to be autonomous, i.e. a central entity will provide each individual sub-detector component with a well defined clock. Data processing of the data collected by a sub-detector needs to be handled by that detector itself and needs to be synchronized with the externally supplied clock. Such an architecture requires that not only digitization but also data compression and data merging is handled by the detector components in order to reduce the necessary bandwidth to acceptable levels. The on-detector electronics will be realized in ASICs and will be directly mounted to the readout cells. These developments will largely profit from the experience gained with the ALICE TRD electronics development [8].

A1.4. Studies for a Ring Imaging Cherenkov detector

The electron RICH detector (RICH1) will provide identification of electrons with γ ≥ 40 corresponding to the momentum range of dielectrons from vector meson decays. The RICH signals can also be used to identify pions and kaons starting from momentum of 5.6 GeV/c. It is positioned at about 1.5 m downstream of the target and consists of a 2.2 m long gas radiator, two arrays of spherical hexagonal mirrors, two photodetector planes and corresponding support structure.

Optical scheme and mirrors

The optical arrangement of the RICH1 detector in vertical and horizontal planes is shown in Fig. A1.6. It consists of two identical spherical mirrors rotated by 120 in the vertical plane. The radius of surface curvature is equal to 450 cm. Each spherical mirror has overall dimensions of 1.75x4.5 m2. The Cherenkov light emitted from charged particles is focused on two focal planes, positioned at 25 cm just after the magnet and at 225 cm before the mirrors.

Our idea is to produce the RICH mirrors from spherical hexagonal 3 mm thick Beryllium plates covered with 0.5 mm glass, which forms an optically perfect mirror surface. The diameter (maximum size) of the hexagons is 60 cm. The diameter of the light point source on the focal plane is less than 0.5 mm. The total radiation length of the mirror (Be + glass) is 1.25% of X0. The weight of one hexagonal plate is about 1.3 kg. The technology of the mirror production exists in Russia.

Photodetectors

The RICH1 detector has two photodetector planes with an area of 3.6 m2 each. It is necessary to detect single photons with the highest possible efficiency and with an optimal photodetector granularity in the region of 7x7 mm2 - 10x10 mm2. Detailed simulations will be performed to define the granularity. The photodetectors should be sensitive to the light in the visible, ultraviolet and vacuum ultraviolet regions.

40

We will investigate the possibility to use an array of photomultipliers as photodetectors. The IHEP group, together with the company Moscow Electrolamp, intend to develop a low-price photomultiplier. To improve the sensitivity of the photomultiplier tubes to UV radiation, the external surface of glass windows of photocathodes will be covered with a transparent film from p-teraphenyl using the technology developed and used in IHEP [19]. The PMTs will be assembled in the photodetector planes using hexagonal packing. Although the geometrical acceptance in this case is equal to 60 % due to 1 mm thickness of the PMT wall compared with 5 mm photomultiplier tube radius, a fraction of more than 90 % of the total photodetector area is assumed to be active. This value will be obtained by using the cone-shaped reflectors from aluminized mylar collecting light on an active part of the photocathode window of each PMT. Therefore the overall detection efficiency for Cherenkov photons is 21 % for photons with λ > 400 nm and 17 % for photons with λ < 400 nm.

Gaseous photodetectors are also considered as an alternative to detect the Cherenkov light. This solution ist cost-effective and copes with any granularity, provided the photocathode is suitably segmented into readout pads [20,21]. Low energy background particles will not reach the photodetector since they are swept away by the dipole magnet upstream. Moreover, the photodetector is shielded from the target by the magnet yoke.

Radiator gas

The basic design feature of the RICH1 detector is the absence of a window between the photodetectors and the gas radiator. Due to this feature we can choose the radiator gas potentially as a mixture of four gases: C4H10, CH4, N2 and He. The He admixture in the radiator leads to hard requirements for the RICH gas system due to high He leakage. The C4H10+N2 gas mixture is much more stable. The variation of the C4H10 admixture allows to obtain radiation thresholds ranging from γthreshold = 16.2 (pure C4H10 radiator) to γthreshold = 41 (pure N2). An admixture of CH4 could be used instead of N2 to increase the radiation length of the radiator.

Fig. A1.6: Optical scheme of RICH1. Left: vertical; right: horizontal

41

A1.5. Recent achievements in R&D on Resistive Plate Chambers

The developments for a RPC solution for CBM proceed currently in two directions: the improvement of the rate capability and of the coverage of large area.

Rate Capability

This part of the programme is pursued mostly by the Coimbra group [22-28]. Of particular relevance for the proposed CBM experiment is the recent extension of the counting rate capabilities of timing RPCs from the current maximum of less than 2kHz/cm2 to 25kHz/cm2, while keeping a time resolution below σ = 100 ps (see figure A1.7). The single-gap counters were made with metal and plastic (ENSITAL) electrodes with a measured resistivity of 109 Ωcm.

Figure A1.7: A time resolution below 100 ps σ is kept essentially unchanged from 2 kHz/cm2 to 27 kHz/cm2. Events were generated by Compton cascades started from an intense 137Cs source. It should be noted that it is know from experience that the time resolution achievable with photons is lower than what is possible in a particle beam.

Coverage of large area and electronics system aspects

One of the main challenges for CBM will be the coverage of about 100 m2 with high performance TOF counters. Other readout models beside single pads will be needed. Towards that end RPC counters with strip readout are developed for application in FOPI [29-31]. Their system behaviour allows to project to the future CBM system.

In figure A1.8 the response of a 6 gap prototype with 4 cm width is shown in terms of time resolution and efficiency. In this setup, a short version (20cm) was tested, which behaves in the same way as the 90 cm long version. The results are obtained with a test beam of minimum ionizing protons. A time resolution well below 100 ps is observed with the efficiency close to 100 %. The contribution of the electronics to the time resolution was determined to only 30 ps.

42

Figure A1.8: Efficiency (left axis) and time resolution (right axis) measured with a 20 cm long prototype (4 cm width) with 6 gaps using minimum ionizing protons

A1.6. The electromagnetic calorimeter system

The calorimeter system of CBM will be used for the identification of leptons and photons. The physics programme requires very effective particle identification for leptons and hadrons, high energy (momentum) and spatial resolution for the reconstruction of short lived heavy particles (J/ψ, D, ρ, ω mesons) via their decay products, very fast response to operate at an unprecedented interaction rate of about 10 MHz, and a highly selective trigger scheme.

Lead-scintillator calorimeter

The calorimeter system is able to measure electrons and photons, providing unique information for lepton (electron and muon) identification and performing a rather accurate time-of-flight analysis. The large angular acceptance of the CBM experiment and the distance needed for the time-of-flight measurement result in a large area calorimeter system located immediately after the wall of resistive plate counters (at least 12 meters from the target). Optimising the cost-to-performance ratio we propose to employ the “shashlik” technology of sampling scintillator-lead structures readout by plastic wavelength shifting fibers, which has been used successfully for the PHENIX, HERA-B and LHCb calorimeters. Adopting this technology for CBM certainly requires considerable modifications.

The energy resolution of the calorimeter will directly affect the sensitivity of our experiment to prompt photons. Good energy resolution is needed to reconstruct π0 and η decays into photon pairs precisely enough to recognize them above the abundant random combinations. Taking into account that the energy range of particles produced in ion-ion collisions at the future GSI accelerator is rather soft, varying from 0.5 - 1 GeV in the outermost region up to 10 - 15 GeV in the innermost region close to the beam pipe, we have to decrease the thickness of lead plates, thus minimising the sampling term in energy resolution. Calorimeter modules built with 0.5 - 1.0 mm thick lead plates have demonstrated already an energy resolution of 5 – 6 %/√E.

At the same time, the high density of incoming particles at CBM – about 1200 particles per heavy ion (Au+Au) central collision – has to be taken into account. Overlapping showers from neighbour tracks could considerably dilute the intrinsic energy resolution of the calorimeter. Figure A1.9 shows the distribution of the distance between the entrance point of electrons (photons) and the closest energetic track in the innermost region of calorimeter located at 12 m from the target, calculated for minimum bias events. The average value of this distribution depends on the distance from the beam pipe as shown in figure A1.10 for minimum bias and

43

central events. The central region of the calorimeter obviously suffers from high occupancy. There are only two possibilities to minimise this effect: decreasing the thickness of scintillator plates (decrease of Moliere radius) or moving the central region of the calorimeter further away from the target, which certainly would increase the overall detector dimensions, complexity and cost.

Lepton identification

Discrimination between e and π in the electromagnetic calorimeter is primarily achieved by the comparison of deposited energy to the measured track momentum. The effectiveness depends on the calorimeter energy resolution as well as on the calorimeter response to hadrons. Dedicated GEANT3 based simulations were used to estimate the e/π discrimination capability of the proposed shashlik calorimeter (25·X0 deep) for single particles within the CBM momentum range. Hadron rejection was calculated for energy deposition cuts corresponding to 90 % electron efficiency. The rejection ranges from at least 30:1 to 100:1 for momentum of 1 - 3 GeV. Further discrimination may be obtained by using shower-shape information (for higher energies) or by time-of-flight information (for lower energies). The calorimeter thickness should also be optimised to find a reasonable compromise, which preserves the energy resolution for high momentum particles while providing valuable rejection for low momentum hadrons.

Discrimination between µ and π may be achieved with a longitudinally segmented calorimeter. This is based on the fact that muon energy depositions are directly proportional to the width of calorimeter segments while pion energy depositions fluctuate significantly. The discriminating power of this method was studied with GEANT3. The hadron rejection ranges from at least 20:1 to 50:1 for momenta of 1 - 3 GeV.

Granularity

The hit density varies significantly (by two orders of magnitude) over the calorimeter surface. It is therefore natural to adopt a variable lateral segmentation in three different cell-size zones following the HERA-B and LHCb experiences. Shashlik technology allows an easy way to build the whole system with the same size basic modules. The required lateral segmentation is achieved dividing scintillator plastic plates into the corresponding number of light isolated pieces and grouping their fibers onto the separate photodetectors:

1. Outer region - single cell modules

2. Middle region - 2x2 cells per module

3. Inner region - 3x3 (or even 4x4) cells per module

The cell repartition between the three zones and the size of basic module should be optimised for the CBM physics programme.

The longitudinal segmentation of the readout within the calorimeter towers is an additional method to discriminate between leptons and hadrons. It also helps to improve the time resolution. The decision on the practicality of longitudinal segmentation depends on the complexity and cost of the additional photodetectors and electronic channels.

44

Figure A1.9 (above): Distance between electron (photon) and closest track in the innermost calorimeter region

Figure A1.10 (right): Average distance between electron (photon) and closest track as a function of the distance from the beam axis. Red symbols: central events. Black symbols: minimum bias events.

Photodetectors

Custom photomultipliers have demonstrated perfect performance with shashlik calorimeters during the last 15 years. However, the relatively high price of these photodetectors seriously limits the granularity of the system, which is particularly important for CBM. Therefore it would be very interesting to consider the new cheaper photodetectors (APDs, multianode PMTs), which might become available in the next 5 years.

R&D programme

Dedicated R&D programme is foreseen to optimise the design of the shashlik-type calorimeter system for the CBM experiment:

1. die stamping technologies for production of thin (0.5 - 1.0 mm) lead tiles;

2. new casting technology for production of thin (0.5 – 1.0 mm) plastic tiles;

3. special chemical treatment of scintillating tile edges to improve the light collection uniformity;

4. detailed Monte-Carlo studies of the calorimeter performance. Optimisation of the lateral and longitudinal granularity;

5. tests of new low-price photodetectors;

6. design of fast electronics.

45

A1.7. CVD Diamond Strip Detectors for TOF measurements and beam monitoring

Poly-Crystalline CVD Diamond Detectors (PC-CVDDD) are radiation-hard and fast particle sensors suitable for operation in primary heavy-ion (HI) beams of intensities up to 109 ions/s [32]. We intend to apply PC-CVDD micro-strip detector(s) of a size of 20x20 mm² and a thickness 100 µm < dD < 300 µm as START (T0) detector(s) for TOF measurements and beam monitoring in the CBM experiment.

The detector(s) will be optimised with respect to an intrinsic time resolution in the order of 50 ps and a count rate capability > 106 ions/s per strip. Presently, a strip and readout pitch of 50 µm at a strip width of 25 µm is discussed. This design allows for operation of the device(s) in different distances from the target including beam focus (active target).

PC-CVDD strip detectors of such a layout have been reliably produced and extensively tested in the past striving for minimum ionising particles tracking near the interaction region in high-energy experiments at high luminosity colliders [33]. Using VA2 electronics, a spatial resolution of 10 µm to 20 µm and a hit efficiency of 86.7% have been achieved with muon beams. GSI, on the other hand, has a long experience with the operation of fast HI diamond detectors placed directly in the beam (start-veto detectors installed by HADES or accelerator beam-diagnostics applications). Using broadband amplifiers developed in house, an intrinsic time resolution well below 50 ps and rate capabilities > 5 ·108/s per channel have been measured.

The availability of experience in both particle detection techniques is advantageous. However, the development of FEE for diamond micro-strip TOF detectors is challenging. TOF ASICs for heavy ions have to cope with complex event structures and to address several design challenges: A fast preamplifier with several GHz bandwidth, a fast discrimination scheme with time walk correction, time stamping with 50 ps resolution, readout and buffering architecture for continuous high-rate data readout. A suitable chip technology (DSM, Bipolar, SiGe) has to be selected.

Figure A1.11:

A CVD diamond beam telescope: Two hybrids carrying diamond strip detectors and readout chips. The 2x2 cm² big sensors of a strip- and readout pitch of 50 µm are centered over the hole in the aluminum frame. The price of the diamond raw material amounts to 12.85 US$ per mm². The metallised detector chip as shown here costs about 5000 EUR.

46

Appendix 2: The superconducting dipole magnet

The requirements imposed onto the CBM magnet are an aperture of 1 x 1 m2 and a field of 1 T confined within 1 m from the target. To meet these requirements, some modifications were introduced to the clamps and yoke with respect to the design presented in the CDR. The current concept is shown in figure A2.1. The magnetic field is calculated using the TOSCA code and implemented in the simulation code. Figure A2.2 shows the y component of the field along the beam axis at x = 0 and various vertical distances from the midplane. A possible technical realisation of the magnet is illustrated in figure A2.3

Figure A2.1: 3D view of the SC magnet Figure A2.2: Calculated vertical field at various distances Y from the mid-plane of the magnet

In order to increase the bending power of the magnet and to reduce the field at the location of the Cherenkov detector, a version with inclined pole shoes has been studied. Figure A2.4 shows a 3D view of the modified magnet and; the corresponding By component is compared in figure A2.5 to that of the standard design.

Figure A2.3: Realisation of the current magnet design. The weight of the iron yoke is 67000 kg. The dimensions of the blocks are chosen such that the weight of one block does not exceed 10000 kg. The coil shape consists of two half-circles (radius = 0,5 m) connected by a 12 cm long straight section. Coil: 2 pancakes, 9 layers, 15 turns. The cross section of the turns is 4,5 x 7 mm2

47

Figure A2.4: 3D view of a magnet with inclined pole shoes

Figure A2.5: By component of the standard magnet (OSDM) and of the inclined magnet (NSDM). Here, z = 0.5 m corresponds to z = 0 in figure A2.2.

Appendix 3: HADES@SIS100/300 The main goal of this study is a simulation of dilepton production in heavy-ion collisions at bombarding energies about 8 AGeV as seen by the HADES detector. The chosen energy is in the lower part of the energy range of bombarding energies that will become available at the future GSI accelerator facilities, i. e. 2 - 45 AGeV.

The primary goal of the HADES project is to investigate various processes contributing to dilepton emission in hot and compressed nuclear matter as created in nucleus-nucleus collisions at bombarding energies between 1 and 2 AGeV. The experimental approach is to measure vector mesons (ρ, ω and eventually φ) via their decay into a pair of leptons, from which their in-medium masses can be reconstructed. The major experimental challenge is to discriminate the penetrating, but very rare leptons from the huge hadronic background which exceeds the electron signal by many orders of magnitude. The HADES detector has been specifically designed to overcome these difficulties.

The dielectron spectrometer HADES at GSI (http://www-hades.gsi.de) has been designed with excellent lepton/hadron discrimination, a mass resolution of ∆M/M = 1 - 2 % and an acceptance of ε = 40 %. In addition, HADES is equipped with a very fast data read-out system as well as a sophisticated 2nd-level trigger system, which both allow to acquire dilepton data at high rates. With these design features, HADES can obtain data with much higher quality and statistical significance than any of the previous dilepton experiments. HADES has been installed at GSI over the last years and in a series of commissioning runs its performance has been checked and gradually improved. In parallel to the work on the hardware, a complete simulation and analysis framework, based on the CERN Root software, has been designed and implemented. This code is by now routinely used to analyse HADES data. Full event simulations are possible and are used to compare with the measured data.

While the CBM experiment will be designed to study heavy-ion collisions starting from 10 AGeV, the energy range from 2 – 10 AGeV, accessible also with the future accelerator

48

facility, can be covered by an upgrade of the existing HADES setup. At present we have started with first simulation studies of the HADES performance measuring the C+C system at 8 AGeV. We have selected the same spectrometer setup as in the experiment C+C at 2 AGeV (NOV01) to be able to compare to results with existing simulation and experimental data. Two simulations have been performed: one to estimate hadron multiplicities and acceptances in the C+C system, and second to get lepton and dilepton acceptance. Events were generated by UrQMD for the hadronic part; for leptons we have used products of the η Dalitz decay provided by the PLUTO generator. 50K events were generated for each source and for both 2 and 8 AGeV. The propagation of particles and the detector response was simulated using a Geant3 based package including the complete HADES geometry and materials. Digitization of the simulated information to the raw data format was done. The resulting digital data were analysed in the same way as experimental ones.

Charged particle multiplicities at 8 A GeV are by a factor of two higher in the forward part of HADES, due to secondary particles created in the material. This probably makes the use of the presently installed TOFINO detector impossible. Without any modification of the spectrometer, the acceptance for pions decreases by 20 – 30 %, for leptons by 10 % and for dileptons by 20 % when going from 2 AGeV to 8 AGeV (see figure A3.1).

Fig.A3.1: Invariant mass spectrum of di-leptons from the η Dalitz decay from C+C at 8 (up) and 2 (down) AGeV, detected in HADES. the number of dilepton pairs is by 20 % lower in the 8 AGeV case (with the same 50K pairs emitted from the source).

Details of the work done up to now can be found via the HADES web page (http://www.hades-gsi.de ->Documents->P. Tlusty, Simulations for HADES at 8 AGeV). The study is ongoing; presently the full dilepton cocktail and the combinatorial background is being simulated. The next steps will include:

• implementation of the planned high granularity inner TOF system based on RPCs, • high momentum resolution analysis using all planes of the MDC chambers, • implementation of a new analysis code with improved lepton identification, • extension of this study to heavier collision systems up to Au+Au.

49

References [1] R. Turchetta et al., Nucl. Inst. Meth. A 458 (2001) 677 [2] M.Winter et al., Proc. of the 8th ICATPP, Como (Italy), October 2003 [3] M. Deveaux et al., Nucl. Inst. Meth. A 512 (2003) 71 [4] G.Deptuch et al., Proc. of the IEEE Nuclear Science Symposium, Portland (Oregon),

September 2003; to be printed in IEEE Trans. Nucl. Sci. (2004). [5] W.Dulinski et al., Proc. of the IEEE Nuclear Science Symposium, Portland (Oregon),

September 2003; to be printed in IEEE Trans. Nucl. Sci. (2004). [6] http://ireswww.in2p3.fr/ires/recherche/capteurs/index.html [7] ALICE Collaboration, Addendum to ALICE Proposal, CERN/LHCC 99-13, LHCC/P3

Addendum 2 (1999) [8] ALICE Transition Radiation Detector, Technical Design Report, CERN/LHCC 2001-021

(2001) [9] A. Andronic et al., IEEE Trans. Nucl. Sci. 48, 1259 [10] A. Wilk, Diploma Thesis, Universität Münster, in preparation. [11] F. Sauli, Nucl. Instr. Meth. A 386 (1997) 531 [12] I. Giomataris et al., Nucl. Instr. Meth. A 376 (1996) 29. [13] A. Oed, Nucl. Instr. Meth. A 449 (2000) 446 F. Bartol et al., J. Phys. III France 6 (1996) 337 [14] C. Garabatos, Yu. Zanevsky, N. Chomeiko, G. Becivenni, et al., INTAS Proposal for

GSI-INTAS Call 2003 [15] J. Wessels et al., Joint Research Acticity 4 of the Integrated Infrastructure Initiative

'Hadron Physics', EU-FP6 (2003) [16] A. Romaniouk, Frascati Physics Series XXV (2002) 93 [17] ATLAS Inner Detector Technical Design Report, CERN/LHCC/97-17, ATLAS TDR 5

(1997) 593 [18] L. Naumann, V. Peshekhonov, et al, INTAS Proposal for GSI-INTAS Call 2003 [19] A. Kozhevnikov et al., Nucl. Instr. and Meth. A 433 (1999) 164. [20] K. Zeitelhack et al., Nucl. Instr. Meth. A 433 (1999) 201 [21] F. Piuz et al., TDR HMPID – Detector for high momentum PID, CERN/LHCC 98-18,

ALICE TDR 1 (1998) [22] P. Fonte and V. Peskov, Nucl. Instr. Meth. A 477 (2002) 17 [23] A. Blanco et al., Nucl. Instr. Meth. A 485 (2002) 328 [24] A. Blanco et al., Nucl. Instr. Meth. A 508 (2003) 63 [25] A. Blanco et al., Nucl. Instr. Meth. A 508 (2003) 70 [27] A. Blanco et al., Nucl. Instr. Meth. A 513 (2003) 8 [28] A. Devismes et al., Nucl. Instr. Meth. A 482 (2002) 179 [29] M. Petrovici et al., Nucl. Instr. Meth. A 487 (2002) 337 [30] M. Petrovici et al., Nucl. Instr. Meth. A 508 (2003) 75 [31] A. Schuettauf et al., GSI annual report (2003) [32] E. Berdermann et al., "The use of CVD-diamond for heavy-ion detection", Diamond and

Related Materials 10 (2001) 1770 [33] D. Meier for the RD42 Collaboration, "A CVD Diamond Beam Telescope for Charged

Particle Tracking", CERN-EP-2001-089